Prediction of acid site evolution and distribution using first-principles kinetic Monte Carlo and Genetic Algorithms: Implications for catalytic hydrotreating of biomass




Journal Title

Journal ISSN

Volume Title



Computational design of multifunctional catalysts and active sites that dynamically respond to changes in reaction conditions is a formidable challenge. Microkinetic models (MKM) are popular for catalytic activity and trend predictions, but the underlying mean field approximation cannot account for the spatial distribution of multiple active sites. In contrast, lattice-based kinetic Monte Carlo (kMC) simulations using a cluster expansion representation of adsorbate−adsorbate interactions can capture the intricacies and complexities of a real catalytic system. I have evaluated the practical advantages of MKM and kMC simulations and found that MKM is faster and more robust when used with limited energetic input information, as it would be common for catalyst screening. The advantages of first-principles kMC simulations are better demonstrated for selective hydrodeoxygenation (HDO) over Ru/TiO2(110) catalysts. My kMC results qualitatively reproduce experimental results and explain the observed effect of co-feeding water on HDO activity. As the water partial pressure increases, the surface coverage of hydroxyl/water increases and the dominant HDO pathway switches from a reverse Mars van Krevelen mechanism to a proton-assisted pathway. There exists no comparable model with the ability to discriminate between metal, oxide and interfacial activity, and assign a dominant catalytic role to each component. While dynamic active site changes under reaction conditions are important, the speciation of active sites is primarily determined by the synthesis conditions. To this end, I developed a genetic algorithm to quickly determine the global minimum energy structure (GMES) of aluminosilicates, and demonstrated a ca. 35% reduction in computational effort compared to a brute force approach. The approaches and outcomes described in this thesis are leading computational catalysis beyond the design of a static active site and towards the ability to develop multifunctional catalysts that dynamically respond to changes in the reactive environment. Such knowledge is necessary for the discovery of highly active and selective catalysts to sustainably produce fuels and chemicals.



Catalysis, kinetic Monte Carlo